Anthropic Restores Claude Fable 5 After Government-Triggered Shutdown

Anthropic brought Claude Fable 5 back online on July 1, ending one of the most dramatic frontier-model rollouts in recent memory. The model originally launched on June 9 alongside Mythos 5, but went dark globally three days later after a U.S. export-control directive triggered by an Amazon-reported cybersecurity safeguard bypass. After a government review, Fable 5 was restored with new classifier safeguards and a proposed framework for scoring jailbreak severity. Mythos 5, however, remains limited to approved U.S.-based organizations, underscoring how regulatory exposure can fragment model availability.

Beyond the shutdown saga, Fable 5 is proving its mettle in real-world benchmarks. The CAIS and Scale AI Labs Remote Labor Index ranked it first among public models, achieving a 16.1% acceptance rate on 240 real remote-work projects across 23 domains. This metric measures whether a client would accept the AI's output as a deliverable. Meanwhile, other AI players made moves: Meta popped 9% on plans to sell excess compute, xAI launched a no-code voice agent builder at $0.05/min, and NVIDIA released faster diffusion language models. The episode signals that frontier AI roadmaps are now inseparable from policy and compliance considerations.